bs_bs_banner
R E V I E W
Drug and Alcohol Review (2015) DOI: 10.1111/dar.12305
Recruiting for addiction research via Facebook LOUISE K. THORNTON1,4, KEITH HARRIS2, AMANDA L. BAKER1, MARTIN JOHNSON3 & FRANCES J. KAY-LAMBKIN1,4 1
Centre for Translational Neuroscience and Mental Health, The University of Newcastle, Newcastle, New South Wales, Australia, 2School of Psychology, University of Queensland, Brisbane, Queensland, Australia, 3School of Psychology, The University of Newcastle, Newcastle, New South Wales, Australia, and 4National Drug and Alcohol Research Centre, University of New South Wales, Sydney, New South Wales, Australia
Abstract Introduction and Aims. This study aimed to examine the feasibility of recruiting participants to addiction research via Facebook. Design and Methods. Participants were recruited via an advertisement on Facebook, a local research register and university psychology courses. Participants completed a self-report survey regarding substance use, history of mental health issues and current psychological distress. Results. The 524 participants recruited via Facebook cost $1.86 per participant; and 418 participants were recruited via more traditional methods. There were significantly fewer women in the Facebook sample compared with the non-Facebook sample (χ2 = 196.61, P < 0.001), but no differences on age. Significantly more Facebook participants reported current use of tobacco (women: Facebook = 57%, non-Facebook = 21%, χ2 = 39.71, P < 0.001; men: Facebook = 62%, non-Facebook = 21%, χ2 = 32.429, P < 0.001) and cannabis (women: Facebook = 26%, nonFacebook = 7%, χ2 = 14.364, P < 0.001; men: Facebook = 46%, non-Facebook = 24%, χ2 = 6.765, P < 0.01). They also reported significantly more harmful use of tobacco [women: F degrees of freedom (d.f.) = 6.07, P < 0.05; men: F(d.f.) = 9.03, P < 0.01] and cannabis [women: F(d.f.) = 11.00, P < 0.01; men: F(d.f.) = 6.40, P < 0.05].The Facebook sample contained a higher percentage of high-severity cannabis users (women: Facebook = 24%, non-Facebook = 4%, χ2 = 18.12, P < 0.001; men: Facebook = 43%, non-Facebook = 16%, χ2 = 10.00, P < 0.01) and reported significantly more severe depressive symptoms [women: F(d.f.) = 26.38, P < 0.001; men: F(d.f.) = 7.44, P < 0.05]. Discussion and Conclusions. Through Facebook, we were able to capture a greater proportion of people with high-severity substance use and mental health issues and were able to capture a greater and more severe range of substance use behaviours. This suggests social networking sites are efficient, cost-effective ways to recruit large numbers of participants, with relevant behaviours and conditions, to addiction research. [Thornton LK, Harris K, Baker AL, Johnson M, Kay-Lambkin FJ. Recruiting for addiction research via Facebook. Drug Alcohol Rev 2015] Key words: social networking site, social media, research subject recruitment, addiction.
Introduction Recruitment of research participants can be difficult and expensive, especially when addressing highly sensitive topics, or trying to recruit low-incidence, highly mobile or stigmatised groups of people, as is often the case in addiction research [1–4]. The Internet, and social networking or social media sites in particular, are potentially efficient and cost-effective ways in which to recruit participants that are currently underutilised by addiction researchers [1–3,5,6].
Social media sites can be defined as Internet sites where users can create a public profile, create a list of other users who they share a connection with (e.g. Facebook friends) and view the content and list of connections made by themselves and others [7]. Use of these types of sites has become one of the most common activities people perform online [8] with many users visiting these sites several times a day [9]. The increasing ease of access to these types of sites via mobile devices has also made it easier for users to access them multiple times a day. The sheer number of
Louise K. Thornton PhD, Post-doctoral Fellow, Keith Harris PhD, Professor, Amanda L. Baker PhD, Professor, Martin Johnson PhD, Senior Lecturer, Frances J. Kay-Lambkin PhD, Associate Professor. Correspondence to Ms Louise Thornton, Centre for Translational Neuroscience and Mental Health, The University of Newcastle, University Drive Callaghan, NSW, 2308, Australia. Tel: +61 2 40335719; Fax: +61 2 4033 5692; E-mail: [email protected] Received 10 July 2014; accepted for publication25 May 2015. © 2015 Australasian Professional Society on Alcohol and other Drugs
2
L. K. Thornton et al.
people who are accessible via social media sites, such as Facebook (http://www.facebook.com) and Twitter (http://www.twitter.com), makes them an attractive recruitment source for researchers [10]. Of particular interest for addiction researchers, is the suggestion that recruitment through social media sites like Facebook can enable the inclusion of low-incidence or hidden population groups [3,4]. Samples of parents of teenagers [11], HIV positive and negative men who have sex with men [12], young adult smokers [3,13], children with life-threatening illnesses and their families [14,15], and low-income populations [16], have also been successfully recruited via Facebook. All of these studies found Facebook to be a cost-effective way (cost per participant in these studies ranged from $1.36 to $8.80) to recruit diverse and large numbers of people often among low-incidence or traditionally hard-toreach populations. O’Conner, Jackson, Goldsmith and Skirton [17] similarly reported that the micro-blogging website Twitter is a cost-effective and efficient means of recruitment. Using a Twitter account to tweet invitations to participate in their research and requesting other Twitter users to retweet their invitation to participate, O’Conner et al. [17] were able to recruit 299 mothers to complete their survey regarding the experiences of older mothers. Online approaches may also reduce bias in response to sensitive topics such as substance use [3,4]. In contrast to face-to-face or telephone interviews that involve disclosure in front of an interviewer, it has been suggested that the social distance and impersonal nature of many online research approaches may reduce the biasing influence of social desirability when asking about sensitive and stigmatised topics [18–20]. Crutzen and Goritz [19], for example, investigated the association between social desirability and self-reported heath risk behaviours in online research. In contrast to offline studies that have found significant relationships between social desirability and reporting of risky behaviours [21], they found that social desirability did not influence reporting of substance use online. They proposed that a possible explanation for this finding could be the perceived privacy of the online setting. Similarly Joinson et al. [22] found that disclosing personal or sensitive information is positively related to degree of anonymity, a factor that online research is well positioned to deliver. Facebook has been used to successfully recruit adolescents to research investigating a wide range of sensitive topics including mental health [23], sexually transmitted infections [2,24], post-traumatic stress disorder [25] and congenital chromosomal defects [26]. Similarly Ramo and Prochaska [3] noted that while they found it particularly difficult to recruit smokers who were unmotivated to quit smoking to their previ© 2015 Australasian Professional Society on Alcohol and other Drugs
ous research, nearly half of their Facebook-recruited participants reported being unmotivated to quit smoking. Research participants can be recruited via social media sites in a number of ways. Using Facebook, as an example, researchers can choose to place details of their study and invitations to participate, on the Facebook pages of existing groups dedicated to the topic of interest. Researchers can also utilise the paid advertising feature of Facebook. Researchers produce an advertisement for their study that appears on the profiles of Facebook users.These advertisements consist of a short title (up to 25 characters), an image and up to 135 characters of text [27]. Using this method, researchers can target their advertisement to a specific population (i.e. have their advertisement only appear on the profiles of certain Facebook users). Facebook advertisements can be targeted by a range of demographic variables (e.g. age, gender, location) as well as by the interests Facebook users list on their profiles. In addition, Facebook displays advertisements on individual Facebook users’ profiles based on their activity on the site such as ‘liking’ a page, in an attempt to display advertisements the user is likely to be most interested in [27]. Facebook is also a useful tool for ‘snowball’ sampling. Brickman Bhutta [28], for example, was able to recruit over 3000 people to an online survey in less than 1 month by utilising exiting networks of Facebook users. To do so, she invited her personal Facebook friends and the administrators of 43 relevant Facebook groups to join a Facebook group related to her research and to share the link to her online survey with their Facebook networks [28]. The current paper describes a secondary analysis of a study that investigated the use of tobacco, alcohol and cannabis, and attitudes towards these substances. The study recruited participants via an advertisement on Facebook, as well as via more traditional methods such as university psychology courses and mail-outs to members of existing research databases (for a full description see [29]). The current study aimed to examine the feasibility of recruiting participants to addiction research via Facebook, and to elucidate the process of using Facebook as a recruitment source for other researchers. With the exception of research conducted by Ramo and colleagues [3,13] among young adult smokers, this is the first study known to examine the utility of Facebook to recruit participants to addiction research and the first to examine the recruitment of adults of all ages to addiction research via Facebook. We aimed to identify the type of people who are likely to participate in addiction research via Facebook and to examine if standardised measures of substance use performed reliably
Recruiting via Facebook
among this group. We were also interested to see if participants recruited via Facebook differed from participants recruited via more traditional recruitment sources on demographic, substance use and psychiatric symptom variables. Based on previous research suggesting that reporting of substance use online may not be influenced as heavily by social desirability as offline reporting [19], we hypothesised that the Facebook sample would be more likely to report past and current tobacco, alcohol or cannabis use, and report more harmful use of these substances as compared with participants recruited in other ways. Similarly, given the sensitive and stigmatising nature of the topic of mental health, we anticipated greater reporting of a history of mental health issues among the Facebook sample and more severe current psychiatric symptoms as compared with the non-Facebook samples. Methods A paid advertisement for the study was placed on Facebook. The advertisement appeared on the profiles of Facebook users over the age of 18 years, who listed their location as Australia, between 13 August and 15 September 2010. The advertisement was not deliberately targeted in any other way. The advertisement included an image of an alcoholic beverage, marijuana leaf and a cigarette and invited people to ‘Go into the draw to win an iPod by telling us what you think about drugs and alcohol’ (see Figure 1). A charge was incurred every time a Facebook user clicked on the advertisement. A daily budget of $30 was set, meaning once enough users had clicked on the advertisement to incur $30 worth of charges, Facebook stopped running the advertisement that day. As a part of their paid advertising service, Facebook provided information regarding numbers of impressions (the number of times the advertisement appeared on a Facebook profile), clicks, click-through rate and cost per click (the amount paid when someone clicked on the study advertisement). If Facebook users clicked on the advertisement, they were taken to an online information statement, and
Figure 1. Facebook advertisement.
3
then on to a self-report questionnaire, hosted by Zoomerang.com. Zoomerang.com provided the time and date of when participants completed the online questionnaire. The self-report questionnaire asked participants about their current and past use of tobacco, alcohol and cannabis, and assessed a range of attitudes and perceptions towards these substances; the survey also asked for their history of mental health issues and measured current psychological distress, as well as basic demographic information. Hazardous and harmful substance use was measured using the Fagerstrom Test for Nicotine Dependence (FTND) [30], Alcohol Use Disorder Identification Test (AUDIT) [31] and Cannabis Use Disorder Identification Test (CUDIT) [32]. Only participants reporting current (within the last 6 months) tobacco, alcohol and/or cannabis use completed the FTND, AUDIT and/or CUDIT. Current psychological distress was measured using the Depression, Anxiety and Stress Scale (DASS-21) [33]. Participants were also recruited separately via a community research database [the Hunter Medical Research Institute’s (HMRI) research register] and first-year psychology courses at the University of Newcastle, New South Wales (NSW), Australia. The HMRI research register is a database of people from the Hunter region in NSW who are interested in contributing to medical research. A paper version of the selfreport questionnaire was posted to 200 registrants over the age of 18 on the HMRI research register. Psychology students over the age of 18 were recruited via an advertisement similar to the Facebook advertisement on the University of Newcastle School of Psychology’s online experimental management system, which allows students to participate in research projects to earn partial course credit. Psychology students completed the online self-report questionnaire hosted by Zoomerang.com. All participants were able to enter into a draw for the chance to win the iPod prize. Analysis Missing values analyses were conducted on the survey data. We used the expectation maximisation (EM) method to replace missing values. EM has been criticised for reducing SE rates, and P values should be interpreted with caution, but EM is still recommended as the simplest and most reasonable approach for imputing missing values [34]. To assess the comparative reliability of testing hazardous and harmful substance use among people recruited via Facebook and other methods, we examined the psychometrics of the three substance use measures (FTND, AUDIT and CUDIT) for each group. Cut-off scores ≥ 8 were chosen as indicators of © 2015 Australasian Professional Society on Alcohol and other Drugs
4
L. K. Thornton et al.
AUDIT, CUDIT total scores and psychiatric symptom measures (DASS-21).
high-severity tobacco, alcohol and cannabis use on the FTND, AUDIT and CUDIT. Scores of 8 or greater on the AUDIT and CUDIT have been shown to be a reliable indicator of hazardous and harmful use and possible alcohol/cannabis dependence while scores of 8 or greater on the FTND indicate very high dependence. We conducted Pearson’s χ2 analyses among participants reporting any substance use, to assess whether the sample groups differed significantly on the proportion of participants who were identified as high-severity. To investigate if, and how, participants recruited via Facebook differed from those recruited via other methods we combined the university student sample with the HMRI sample for comparisons with the Facebook sample. We compared our Facebook and non-Facebook sample groups on the key demographic factors of sex and age. A one-way analysis of variance (anova) was conducted to compare participant ages across recruitment sources. A Pearson’s χ2 analysis was used to compare sex among the samples. Similarly, the percentages of participants reporting current (within the last week) and past use of tobacco, alcohol and cannabis, and endorsing a history of mental health issues were compared between the Facebook and nonFacebook samples using Pearson’s chi-square analyses. Next we conducted general linear modelling to test how the sample groups compared on the FTND,
Results The study’s Facebook advertisement had 4 106 729 impressions over the course of 1 month.This resulted in a total of 2220 clicks (0.054% click-through rate), of whom 553 went on to complete the self-report questionnaire. Twenty-nine (5.24%) participants were excluded because of excessive missing values leaving a total of 524 participants. Over the course of the campaign, the cost-per-click ranged between $0.30 and $0.63 (mean = $0.44). Overall, the campaign cost $975.83 or $1.86 per participant recruited via Facebook. In addition to the participants recruited via Facebook, 115 HMRI research registrants and 319 psychology students returned the self-report questionnaire completed. Seven (6.08%) HMRI registrants and 9 (2.82%) psychology students were excluded because of excessive missing values, leaving a total of 418 nonFacebook participants. Participant characteristics are displayed in Table 1. We first compared the psychology student and HMRI samples on the FTND, AUDIT, CUDIT and DASS21 subscales, finding no statistically significant
Table 1. Participant characteristics Facebook
Non-Facebook
Male N Age Range Mean (SD) ATSI Substance use Ever smoked Smoked in last week Ever drank Drank in last week Ever used cannabis Used cannabis last week History of mental health issues Depression Anxiety Psychotic disorder Eating disorder Other
352
Female 172
Male 89
Female 329
18–73 30.68 (12.44) % 2.8
18–62 32.72 (12.73) % 4.1
18–86 33.16 (18.87) % 3.4
18–80 29.44 (14.98) % 3.3
88.6 61.6 85.8 75.0 65.1 46.3
77.9 56.8 89 58.6 58.1 26.0
68.5*** 21.7*** 95.5* 72.0 42.7*** 23.7**
51.4*** 21.4*** 93.9* 67.1 35.6*** 7.0***
20.5 12.5 3.2 0.0 6.5
36.6 20.3 1.8 3.5 11.6
20.2 10.1 1.1 1.1 3.4
19.6*** 14.4 0.9 4.3 3.0***
***P < 0.001; **P < 0.01; *P < 0.05. All characteristics are self-reported. ‘Other’ mental health issues listed by participants were primarily bipolar disorder, attention deficit/hyperactivity disorder and borderline personality disorder. ATSI, Aboriginal and Torres Strait Islanders. © 2015 Australasian Professional Society on Alcohol and other Drugs
Recruiting via Facebook
5
Table 2. Drug use scale psychometrics among participants reporting tobacco, alcohol or cannabis use Measure/group AUDIT Facebook Non-Facebook FTND Facebook Non-Facebook CUDIT Facebook Non-Facebook
N
Range
Mean
SD
Alpha
S–B
455 394
0–38 0–30
9.13 6.84
7.52 5.41
0.87 0.80
0.85 0.80
446 230
0–10 0–9
2.45 1.26
2.59 1.63
0.82 0.78
0.91 0.86
329 155
0–39 0–29
8.35 4.23
8.34 4.23
0.88 0.84
0.86 0.90
Facebook = Facebook sample, non-Facebook = university student and community volunteers. Alpha, Cronbach’s alpha; AUDIT, Alcohol Use Disorders Identification Test; CUDIT, Cannabis Use Disorders Identification Test; FTND, Fagerstrom Test for Nicotine Dependence; S–B, Spearman–Brown prophecy coefficient.
Table 3. Sample comparisons on proportions of substance using participants reporting high-risk substance use Women Measure AUDIT (n = 849) FTND (n = 676) CUDIT (n = 484)
Men χ2
Facebook (%) Non-Facebook (%) 32.7 6.0 24.0
32.7 1.8 4.3
φ
Facebook (%) Non-Facebook (%)
0.00 0.00 3.76* 0.11 18.12*** 0.29
55.3 4.2 42.8
47.1 1.6 15.8
χ2
φ
1.81 .07 0.90 .05 10.00** .19
*P = 0.05; **P < 0.01; ***P < 0.001. Facebook = participants recruited via Facebook, non-Facebook = psychology student and community participants. AUDIT, Alcohol Use Disorders Identification Test; CUDIT, Cannabis Use Disorders Identification Test; FTND, Fagerstrom Test for Nicotine Dependence.
differences (P > 0.05). We therefore concluded that the two groups were comparable and combined them as a non-Facebook sample. Table 2 displays the ranges, measures of centrality, and internal reliability values for each of the three substance use measures. We report both Cronbach’s α and Spearman–Brown prophecy coefficient (a measure of split-half reliability) to provide more comprehensive information on scale performance. The Facebook sample was somewhat larger, and therefore it is not surprising that a broader range of total scores was found in that group. Although all measures demonstrated adequate internal reliability (i.e. Cronbach’s α ≥ 0.70), for both samples, the Facebook sample showed small improvements in internal reliability in five of six comparisons. The mostly improved internal reliability levels, for the Facebook sample, are likely to have been influenced by the broader range of total scores on the three measures. As can be seen in Table 1, the Facebook and nonFacebook samples differed on sex, χ2 (N = 942, degrees of freedom = 1) = 196.61, P < 0.001, Phi = 0.46, but did not differ significantly on age. For this reason, subsequent analyses were conducted separately for
males and females. Significantly more male and female Facebook participants reported current and past use of tobacco and cannabis, while significantly more nonFacebook participants reported past use of alcohol; however, the samples did not differ on current use of alcohol. Women in the Facebook sample were more likely to report a history of depression, and ‘other’ mental health issues than women in the non-Facebook sample. Table 3 shows that the Facebook sample generally captured greater percentages of high-severity users of each substance, when compared with the nonFacebook sample, for each sex. However, only cannabis use showed statistically significant differences, for each sex. Lastly, we conducted general linear modelling to determine if the Facebook sample differed from the non-Facebook sample on FTND, AUDIT and CUDIT total scores and DASS21 subscales. As the Facebook and non-Facebook samples were previously found to differ on sex, to better understand how these samples compared on our key variables of interest, we conducted the following analyses separately by sex. Multivariate anova showed significant group effects on the © 2015 Australasian Professional Society on Alcohol and other Drugs
6
L. K. Thornton et al.
Table 4. Results of MANOVAs comparing Facebook and non-Facebook samples on drug use and psychiatric symptoms Women Facebook (n = 89) Variable Drug use AUDIT FTND CUDIT DASS21 Depression Anxiety Stress
Men
Non-Facebook (n = 99)
Mean
SD
Mean
SD
7.82 2.37 6.48
6.99 2.50 7.28
8.67 1.57 3.82
5.40 1.97 3.14
5.61 3.55 5.50
4.75 3.83 4.26
3.54 2.80 4.54
4.02 3.51 3.80
Facebook (n = 219) F
Non-Facebook (n = 35)
d
Mean
SD
Mean
SD
F
d
0.87 6.07* 11.00**
0.14 0.36 0.48
10.90 2.54 9.18
7.65 2.62 8.40
10.80 1.17 5.43
6.78 1.54 6.21
0.01 9.03** 6.40*
0.01 0.55 0.46
26.38*** 4.88**** 6.60*
0.48 0.21 0.24
5.10 3.43 5.34
4.34 3.28 3.80
3.71 2.88 4.38
4.18 3.59 3.84
7.44* 1.96 4.46****
0.32 0.16 .25
*P < 0.05; **P < 0.01; ***P < 0.001 (Bonferroni corrected, α/3); ****P < 0.05, (uncorrected). Facebook = participants recruited via Facebook, non-Facebook = psychology student and community participants Only participants who reported use of tobacco, alcohol and/or cannabis in the past 6 months completed the FTND, AUDIT and/or CUDIT. AUDIT, Alcohol Use Disorders Identification Test; CUDIT, Cannabis Use Disorders Identification Test; DASS, Depression Anxiety Stress Scales; FTND, Fagerstrom Test for Nicotine Dependence; manova, multivariate analysis of variance.
substance use and DASS-21 subscales (P < 0.001), for each sex. As the sample groups showed statistically significant differences on these combined variables, we conducted follow-up anovas to assess each measure separately. Table 4 illustrates that for both men and women, the Facebook sample did report significantly more harmful use of tobacco and cannabis. However, the samples did not differ on reported harmfulness of alcohol use. The Facebook sample reported significantly more severe depressive symptoms. Women, recruited through Facebook, also reported significantly higher stress. Discussion This study is one of the first to examine and describe the process of recruiting participants to addiction research via Facebook. At a cost of $1.86 per participant it demonstrates that recruiting via social media sites is a potentially efficient and cost-effective method for recruiting large numbers of participants, including traditionally hard-to-access groups of people, to addiction research. This study set out to identify the type of person who might participate in addiction research via Facebook. We found that people of a wide range of ages, but mostly men, responded to our Facebook advertisement. With a mean age of 31, results show that this approach may not only be useful for recruiting young people to research, but may also be helpful when aiming to recruit older adults as well. We also found that a large proportion of Facebook participants reported tobacco © 2015 Australasian Professional Society on Alcohol and other Drugs
and cannabis use. However, future studies should examine the relationship between self-reported substance use (as reported via Facebook, for example), with objective measures of substance use (e.g. bloods, urine, collateral reports) to establish the reliability of reporting via this source. This research suggests, however, that social media sites, such as Facebook, may be an effective way to gain access to large numbers of substance users who are willing to participate in research. This study also suggests Facebook may be a useful source for recruiting participants with mental health issues. Previous research examining online recruitment and data collection suggests this approach may reduce bias in response to sensitive topics such as substance use and mental health [19]. We therefore hypothesised that compared with participants recruited via more traditional methods, more participants recruited via Facebook would report use of tobacco, alcohol or cannabis and a history of a mental health issues. Similarly, we also hypothesised that these participants would report more harmful use of these substances and more severe psychiatric symptoms. These hypotheses were, for the most part, supported by the current results. Greater percentages of Facebook participants reported past and current use of tobacco and cannabis and a history of depression. Current use of alcohol did not differ between the groups. Similarly, significantly more Facebook participants reported high-severity cannabis use. Mean FTND and CUDIT scores were also significantly higher among Facebook participants, indicating more harmful use of tobacco and cannabis among
Recruiting via Facebook
Facebook participants. While all DASS21 subscale scores were in the expected direction, indicating greater psychiatric distress among participants recruited via Facebook, only depressive symptoms and stress among women were significantly more severe in the Facebook sample compared with the non-Facebook sample. These results suggest that while samples recruited via social media sites such as Facebook might differ from those recruited via more traditional methods, these differences highlight the potential usefulness of this approach for addiction research. The results also question the utility of combining samples recruited in these different ways, as they appear not to be from the same population. Recruiting online and via social media sites may allow researchers to easily access larger numbers of high-severity substance users, and those with comorbid mental health issues, than is usually possible via traditional methods. A number of criticisms often directed at studies employing Internet recruitment methods could also be levelled at the current study. Concerns are often raised regarding the quality of data collected in Internet studies. In the current study similar percentages of participants were excluded from all three recruitment sources because they did not adequately complete the self-report questionnaire (Facebook: 5.24%; HMRI: 6.08%; psychology students: 2.82%).This suggests that it may be possible to collect equivalent quality data via Facebook in comparison with other recruitment sources. Yet, this study did not directly test the genuineness of participant response; that is, the survey did not make use of questions to elucidate response truthfulness, and did not control for participant response patterns (e.g. always responding A in a multiple choice question). Future research in this area would benefit by directly testing the genuineness of participants responses across recruitment methods. Another limitation of the current study was that because comparing Facebook with other recruitment methods was not an aim of the original study, we did not track the costs associated with the other recruitment methods used. Future research should aim to directly compare the cost-effectiveness of various recruitment methods. Additionally, it should be noted that the samples used as a comparison with the Facebook sample, psychology students and volunteers from community research databases are unlikely to be fully representative of the general population. However, these two groups are common sources of participants in behavioural research [35]. A common criticism of the use of Internet advertising in research is the frequently low click-through rate [2]. In the current study, the click-through rate was 0.054%. Similarly, low click-through rates have been
7
found in other studies using comparable methods [2,3]. A number of researchers have been critical of these low click-through rates, comparing click-through rates with response rates of traditional recruitment methods such as mail-outs. This is an inappropriate comparison with infer, as mail-outs directly approach people to participate in research. A more appropriate comparison with make would be with the recruitment of participants through indirect or passive advertising via traditional media such as TV, radio and print media. The Internet is a quickly changing arena, and recent changes within this sphere may affect the continuing relevance of the current findings. In this study, and previous studies, it is proposed that the greater reporting on sensitive topics observed may have been due to the perceived privacy associated with online research. However, starting in June 2013, Edward Snowden revealed information that showed that many governments, including Australia, are conducting large-scale digital surveillance on their own citizens, and those of other countries [36,37]. It is unclear if events have eroded the perception of anonymity and privacy online to a point where social desirability biases would now influence participants responses to sensitive topics more [36]. Other changes include those involving the way in which advertising campaigns can be developed and disseminated on Facebook. Facebook has made new tools available that allow advertisers to target particular audiences more specifically and provide detailed information about the potential reach of advertisements [27]. For example advertisements can be targeted according to a range of users’ behaviours such as: online spending activity, the device or programme they use to access Facebook, and even if their Facebook activity suggests they are planning to travel or have recently returned from a trip [27]. Despite these limitations, this paper highlights the potential of social media sites, such as Facebook as a recruitment source for addiction research. It is important to note, however, that not all Facebook advertising campaigns are as successful as the campaign employed in the current study, and not all research topics and questions will be appropriate for this method of recruitment. The current study, for example, gave participants the chance to win an iPod, which may have contributed significantly to the success of the campaign. Indeed, in a meta-analysis of online surveys Goritz [38] found that offering an incentive significantly increased the odds of a person responding to, and completing, an online survey. The current study also likely oversampled substance users as a result of the advertising procedure used by Facebook, which meant the advertisement would have been more likely to appear on the pages of people whose activity contained mentions of drugs and alcohol [39]. © 2015 Australasian Professional Society on Alcohol and other Drugs
8
L. K. Thornton et al.
Conclusion Facebook, and other social media sites, have the potential to play an important role in accessing substance users and engaging them with research and potentially treatment, as well as increasing our understanding of substance use and addictive behaviours. This research found that large numbers of cannabis users were willing to participate in research via Facebook. As computerdelivered cannabis use treatment has been found to be effective [40], Facebook and other social media sites could be an effective way to access and engage cannabis users with effective treatment. These sites could also be an effective way to directly reach and engage young people with addiction treatment and research. This approach could address the potential barriers of obtaining parental consent and fear that parents will find out about any substance use the young person reports. Finally, social media sites, like Facebook could also facilitate rapid research regarding new and emerging substances. Acknowledgements This research was supported by Dr Thornton’s Australian Postgraduate Award. References [1] Balfe M, Doyle F, Conroy R. Using Facebook to recruit young adults for qualitative research projects: how difficult is it? Comput Inform Nurs 2012;30:511–15. [2] Fenner Y, Garland SM, Moore EE, et al. Web-based recruiting for health research using a social networking site: an exploratory study. J Med Internet Res 2012;14: e20. [3] Ramo DE, Prochaska JJ. Broad reach and targeted recruitment using Facebook for an online survey of young adult substance use. J Med Internet Res 2012;14:e28. [4] Temple EC, Brown RF. A comparison of Internet-based participant recruitment methods: engaging the hidden population of cannabis users in research. J Res Pract 2011;7(2), Article D2. Retrieved [16th June 2015], from http://jrp.icaap.org/index.php/jrp/article/view/288/247 [5] Deady M, Kay-Lambkin F, Thornton L, Baker A, Teesson M. Social influence, addictions and the internet: the potential of Web 2.0 technologies in enhancing treatment for alcohol/other drug use problems. Addiction Research and Therapy 2012;8:2. [6] Freeman B, Chapman S. Gone viral? Heard the buzz? A guide for public health practitioners and researchers on how Web 2.0 can subvert advertising restrictions and spread health information. J Epidemiol Community Health 2008; 62:778–82. [7] Boyd DM, Ellison NB. Social network sites: definition, history and scholarship. J Comput-Mediat Comm 2008; 13:210–30. [8] Alexa.com. 2014 Alexa.com. Available at: http://www .alexa.com/siteinfo/facebook.com (accessed April 2014). © 2015 Australasian Professional Society on Alcohol and other Drugs
[9] Hamptom K, Sessions G, Rainie L, Purcell K. Social networking sites and our lives. Washington D.C.: PEW Internet & American Life Project, 2011. [10] Farmer AD, Bruckner Holt CEM, Cook MJ, Hearing SD. Social networking sites: a novel portal for communication. Postgrad Med J 2009;85:455–9. [11] Gilligan C, Kypri K, Bourke J. Social networking versus Facebook advertising to recruit survey respondents: a quasi-experimental study. JMIR Res Protoc 2014;3: e48. [12] Hernandez-Romieu AC, Sullivan PS, Sanchez TH, et al. The comparability of men who have sex with men recruited from venue–time–space sampling and Facebook: a cohort study. JMIR Res Protoc 2014;3:e37. [13] Ramo DE, Rodriguez TM, Chavez K, Sommer MJ, Prochaska JJ. Facebook recruitment of young adult smokers for a cessation trial: methods, metrics and lessons learned. Internet Interv 2014;1:58–64. [14] Akard TF, Wray S, Gilmer MJ. Facebook advertisements recruit parents of children with cancer for an online survey of web-based research preferences. Cancer Nurs 2015;38: 155–61. [15] Schumacher KR, Stringer KA, Donohue JE, et al. Social media methods for studying rare diseases. Pediatrics 2014; 133:e1345–53. [16] Lohse B, Wamboldt P. Purposive Facebook recruitment endows cost-effective nutrition education program evaluations. JMIR Res Protoc 2013;2:e27. [17] O’Conner A, Jackson L, Goldsmith L, Skirton H. Can I get a retweet please? Health research recruitment and the Twittersphere. J Adv Nurs 2014;70:599–609. [18] Newman JC, Des Jarlais DC, Turner CF, Gribble J, Cooley P, Paone D. The differential effects of face-to-face and computer interview modes. Am J Public Health 2002;92: 294–7. [19] Crutzen R, Göritz AS. Social desirability and self-reported health risk behaviors in web-based research: three longitudinal studies. BMC Public Health 2010;10:720. [20] Joinson AN. Knowing me, knowing you: reciprocal selfdisclosure in Internet-based surveys. Cyberpsychol Behav 2001;4:587–91. [21] Kogan N. Risk taking: a study in cognition and personality. New York: Holt, Rinehart &Winston, 1964. [22] Joinson AN, Woodley A, Reips UD. Personalization, authentication and self-disclosure in self-administered Internet surveys. Comput Human Behav 2007;23:275– 85. [23] Ellis LA, Collin P, Davenport TA, Hurley PJ, Burns JA, Hickie IB. Young men, mental health, and technology: implications for service design and delivery in the digital age. J Med Internet Res 2012;14:e160. [24] Hilton S, Smith E. ‘I thought cancer was one of those random things. I didn’t know cancer could be caught . . .’: adolescent girls’ understandings and experiences of the HPV programme in the UK. Vaccine 2011;29:4409–15. [25] Chu JL, Snider CE. Use of a social networking web site for recruiting Canadian youth for medical research. J Adolesc Health 2013;52:792–4. [26] Close S, Smaldone A, Fennoy I, Reame N, Grey M. Using information technology and social networking for recruitment of research participants: experience from an exploratory study of pediatric Klinefelter syndrome. J Med Internet Res 2013;15:e48. [27] Facebook 2013 Facebook.com. Available at: http://www .facebook.com (accessed January 2013).
Recruiting via Facebook
[28] Brickman Bhutta C. Not by the book: Facebook as a sampling frame. Sociol Methods Res 2012;41:57–88. [29] Thornton LK, Baker AL, Johnson MP, Lewin TJ. Perceived risk associated with tobacco, alcohol and cannabis use among people with and without psychotic disorders. Addict Behav 2013;38:2246–51. [30] Heatherton TF, Kozlowski LT, Frecker RC, Fagerstrom KO. The Fagerstrom Test for Nicotine Dependence: a revision of the Fagerstrom tolerance questionnaire. Br J Addict 1991;86:1119–27. [31] Saunders JB, Aasland OG, Babor TF, de la Fuente JR, Grant M. Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption—II. Addiction 1993;88:791–804. [32] Adamson SJ, Sellman JD. A prototype screening instrument for cannabis use disorder: the Cannabis Use Disorders Identification Test (CUDIT) in an alcohol-dependent clinical sample. Drug Alcohol Rev 2003;22:309–15. [33] Lovibond SH, Lovibond PF. Manual for the depression anxiety stress scales, 2nd edn. Sydney: Psychology Foundation, 1995.
9
[34] Tabachnick BG, Fidell LS. Using multivariate statistics, 5th edn. Boston: Pearson Education, 2007. [35] Samuels D, Zucco Cr. Using Facebook as a Subject Recruitment Tool for Survey-Experimental Research (August 2013). 2013. Available at SSRN: http://ssrn .com/abstract=2101458 or http://dx.doi.org/10.2139/ssrn .2101458. [36] Lyon D. Surveillance, Snowden, and big data: capacities, consequences, critique. Big Data & Soc 2014;July– September:1–13. [37] Dorling P. Top-secret documents reveal Australia targeting Indonesia, South Pacific mobile phone networks. The Sydney Morning Herald. 2015;2015. [38] Goritz AS. Incentives in web studies: methodological issues and a review. Int J of Internet Science 2006;1:58–70. [39] Facebook. 2014. About Facebook ads. Available at: https:// www.facebook.com/about/ads/ (accessed 16 June 2015). [40] Kay-Lambkin F, Baker A, Lewin T, Carr V. Computerbased psychological treatment for comorbid depression and problematic alcohol and/or cannabis use: a randomized controlled trial of clinical efficacy. Addiction 2009;104: 378–88.
© 2015 Australasian Professional Society on Alcohol and other Drugs
R E V I E W
Drug and Alcohol Review (2015) DOI: 10.1111/dar.12305
Recruiting for addiction research via Facebook LOUISE K. THORNTON1,4, KEITH HARRIS2, AMANDA L. BAKER1, MARTIN JOHNSON3 & FRANCES J. KAY-LAMBKIN1,4 1
Centre for Translational Neuroscience and Mental Health, The University of Newcastle, Newcastle, New South Wales, Australia, 2School of Psychology, University of Queensland, Brisbane, Queensland, Australia, 3School of Psychology, The University of Newcastle, Newcastle, New South Wales, Australia, and 4National Drug and Alcohol Research Centre, University of New South Wales, Sydney, New South Wales, Australia
Abstract Introduction and Aims. This study aimed to examine the feasibility of recruiting participants to addiction research via Facebook. Design and Methods. Participants were recruited via an advertisement on Facebook, a local research register and university psychology courses. Participants completed a self-report survey regarding substance use, history of mental health issues and current psychological distress. Results. The 524 participants recruited via Facebook cost $1.86 per participant; and 418 participants were recruited via more traditional methods. There were significantly fewer women in the Facebook sample compared with the non-Facebook sample (χ2 = 196.61, P < 0.001), but no differences on age. Significantly more Facebook participants reported current use of tobacco (women: Facebook = 57%, non-Facebook = 21%, χ2 = 39.71, P < 0.001; men: Facebook = 62%, non-Facebook = 21%, χ2 = 32.429, P < 0.001) and cannabis (women: Facebook = 26%, nonFacebook = 7%, χ2 = 14.364, P < 0.001; men: Facebook = 46%, non-Facebook = 24%, χ2 = 6.765, P < 0.01). They also reported significantly more harmful use of tobacco [women: F degrees of freedom (d.f.) = 6.07, P < 0.05; men: F(d.f.) = 9.03, P < 0.01] and cannabis [women: F(d.f.) = 11.00, P < 0.01; men: F(d.f.) = 6.40, P < 0.05].The Facebook sample contained a higher percentage of high-severity cannabis users (women: Facebook = 24%, non-Facebook = 4%, χ2 = 18.12, P < 0.001; men: Facebook = 43%, non-Facebook = 16%, χ2 = 10.00, P < 0.01) and reported significantly more severe depressive symptoms [women: F(d.f.) = 26.38, P < 0.001; men: F(d.f.) = 7.44, P < 0.05]. Discussion and Conclusions. Through Facebook, we were able to capture a greater proportion of people with high-severity substance use and mental health issues and were able to capture a greater and more severe range of substance use behaviours. This suggests social networking sites are efficient, cost-effective ways to recruit large numbers of participants, with relevant behaviours and conditions, to addiction research. [Thornton LK, Harris K, Baker AL, Johnson M, Kay-Lambkin FJ. Recruiting for addiction research via Facebook. Drug Alcohol Rev 2015] Key words: social networking site, social media, research subject recruitment, addiction.
Introduction Recruitment of research participants can be difficult and expensive, especially when addressing highly sensitive topics, or trying to recruit low-incidence, highly mobile or stigmatised groups of people, as is often the case in addiction research [1–4]. The Internet, and social networking or social media sites in particular, are potentially efficient and cost-effective ways in which to recruit participants that are currently underutilised by addiction researchers [1–3,5,6].
Social media sites can be defined as Internet sites where users can create a public profile, create a list of other users who they share a connection with (e.g. Facebook friends) and view the content and list of connections made by themselves and others [7]. Use of these types of sites has become one of the most common activities people perform online [8] with many users visiting these sites several times a day [9]. The increasing ease of access to these types of sites via mobile devices has also made it easier for users to access them multiple times a day. The sheer number of
Louise K. Thornton PhD, Post-doctoral Fellow, Keith Harris PhD, Professor, Amanda L. Baker PhD, Professor, Martin Johnson PhD, Senior Lecturer, Frances J. Kay-Lambkin PhD, Associate Professor. Correspondence to Ms Louise Thornton, Centre for Translational Neuroscience and Mental Health, The University of Newcastle, University Drive Callaghan, NSW, 2308, Australia. Tel: +61 2 40335719; Fax: +61 2 4033 5692; E-mail: [email protected] Received 10 July 2014; accepted for publication25 May 2015. © 2015 Australasian Professional Society on Alcohol and other Drugs
2
L. K. Thornton et al.
people who are accessible via social media sites, such as Facebook (http://www.facebook.com) and Twitter (http://www.twitter.com), makes them an attractive recruitment source for researchers [10]. Of particular interest for addiction researchers, is the suggestion that recruitment through social media sites like Facebook can enable the inclusion of low-incidence or hidden population groups [3,4]. Samples of parents of teenagers [11], HIV positive and negative men who have sex with men [12], young adult smokers [3,13], children with life-threatening illnesses and their families [14,15], and low-income populations [16], have also been successfully recruited via Facebook. All of these studies found Facebook to be a cost-effective way (cost per participant in these studies ranged from $1.36 to $8.80) to recruit diverse and large numbers of people often among low-incidence or traditionally hard-toreach populations. O’Conner, Jackson, Goldsmith and Skirton [17] similarly reported that the micro-blogging website Twitter is a cost-effective and efficient means of recruitment. Using a Twitter account to tweet invitations to participate in their research and requesting other Twitter users to retweet their invitation to participate, O’Conner et al. [17] were able to recruit 299 mothers to complete their survey regarding the experiences of older mothers. Online approaches may also reduce bias in response to sensitive topics such as substance use [3,4]. In contrast to face-to-face or telephone interviews that involve disclosure in front of an interviewer, it has been suggested that the social distance and impersonal nature of many online research approaches may reduce the biasing influence of social desirability when asking about sensitive and stigmatised topics [18–20]. Crutzen and Goritz [19], for example, investigated the association between social desirability and self-reported heath risk behaviours in online research. In contrast to offline studies that have found significant relationships between social desirability and reporting of risky behaviours [21], they found that social desirability did not influence reporting of substance use online. They proposed that a possible explanation for this finding could be the perceived privacy of the online setting. Similarly Joinson et al. [22] found that disclosing personal or sensitive information is positively related to degree of anonymity, a factor that online research is well positioned to deliver. Facebook has been used to successfully recruit adolescents to research investigating a wide range of sensitive topics including mental health [23], sexually transmitted infections [2,24], post-traumatic stress disorder [25] and congenital chromosomal defects [26]. Similarly Ramo and Prochaska [3] noted that while they found it particularly difficult to recruit smokers who were unmotivated to quit smoking to their previ© 2015 Australasian Professional Society on Alcohol and other Drugs
ous research, nearly half of their Facebook-recruited participants reported being unmotivated to quit smoking. Research participants can be recruited via social media sites in a number of ways. Using Facebook, as an example, researchers can choose to place details of their study and invitations to participate, on the Facebook pages of existing groups dedicated to the topic of interest. Researchers can also utilise the paid advertising feature of Facebook. Researchers produce an advertisement for their study that appears on the profiles of Facebook users.These advertisements consist of a short title (up to 25 characters), an image and up to 135 characters of text [27]. Using this method, researchers can target their advertisement to a specific population (i.e. have their advertisement only appear on the profiles of certain Facebook users). Facebook advertisements can be targeted by a range of demographic variables (e.g. age, gender, location) as well as by the interests Facebook users list on their profiles. In addition, Facebook displays advertisements on individual Facebook users’ profiles based on their activity on the site such as ‘liking’ a page, in an attempt to display advertisements the user is likely to be most interested in [27]. Facebook is also a useful tool for ‘snowball’ sampling. Brickman Bhutta [28], for example, was able to recruit over 3000 people to an online survey in less than 1 month by utilising exiting networks of Facebook users. To do so, she invited her personal Facebook friends and the administrators of 43 relevant Facebook groups to join a Facebook group related to her research and to share the link to her online survey with their Facebook networks [28]. The current paper describes a secondary analysis of a study that investigated the use of tobacco, alcohol and cannabis, and attitudes towards these substances. The study recruited participants via an advertisement on Facebook, as well as via more traditional methods such as university psychology courses and mail-outs to members of existing research databases (for a full description see [29]). The current study aimed to examine the feasibility of recruiting participants to addiction research via Facebook, and to elucidate the process of using Facebook as a recruitment source for other researchers. With the exception of research conducted by Ramo and colleagues [3,13] among young adult smokers, this is the first study known to examine the utility of Facebook to recruit participants to addiction research and the first to examine the recruitment of adults of all ages to addiction research via Facebook. We aimed to identify the type of people who are likely to participate in addiction research via Facebook and to examine if standardised measures of substance use performed reliably
Recruiting via Facebook
among this group. We were also interested to see if participants recruited via Facebook differed from participants recruited via more traditional recruitment sources on demographic, substance use and psychiatric symptom variables. Based on previous research suggesting that reporting of substance use online may not be influenced as heavily by social desirability as offline reporting [19], we hypothesised that the Facebook sample would be more likely to report past and current tobacco, alcohol or cannabis use, and report more harmful use of these substances as compared with participants recruited in other ways. Similarly, given the sensitive and stigmatising nature of the topic of mental health, we anticipated greater reporting of a history of mental health issues among the Facebook sample and more severe current psychiatric symptoms as compared with the non-Facebook samples. Methods A paid advertisement for the study was placed on Facebook. The advertisement appeared on the profiles of Facebook users over the age of 18 years, who listed their location as Australia, between 13 August and 15 September 2010. The advertisement was not deliberately targeted in any other way. The advertisement included an image of an alcoholic beverage, marijuana leaf and a cigarette and invited people to ‘Go into the draw to win an iPod by telling us what you think about drugs and alcohol’ (see Figure 1). A charge was incurred every time a Facebook user clicked on the advertisement. A daily budget of $30 was set, meaning once enough users had clicked on the advertisement to incur $30 worth of charges, Facebook stopped running the advertisement that day. As a part of their paid advertising service, Facebook provided information regarding numbers of impressions (the number of times the advertisement appeared on a Facebook profile), clicks, click-through rate and cost per click (the amount paid when someone clicked on the study advertisement). If Facebook users clicked on the advertisement, they were taken to an online information statement, and
Figure 1. Facebook advertisement.
3
then on to a self-report questionnaire, hosted by Zoomerang.com. Zoomerang.com provided the time and date of when participants completed the online questionnaire. The self-report questionnaire asked participants about their current and past use of tobacco, alcohol and cannabis, and assessed a range of attitudes and perceptions towards these substances; the survey also asked for their history of mental health issues and measured current psychological distress, as well as basic demographic information. Hazardous and harmful substance use was measured using the Fagerstrom Test for Nicotine Dependence (FTND) [30], Alcohol Use Disorder Identification Test (AUDIT) [31] and Cannabis Use Disorder Identification Test (CUDIT) [32]. Only participants reporting current (within the last 6 months) tobacco, alcohol and/or cannabis use completed the FTND, AUDIT and/or CUDIT. Current psychological distress was measured using the Depression, Anxiety and Stress Scale (DASS-21) [33]. Participants were also recruited separately via a community research database [the Hunter Medical Research Institute’s (HMRI) research register] and first-year psychology courses at the University of Newcastle, New South Wales (NSW), Australia. The HMRI research register is a database of people from the Hunter region in NSW who are interested in contributing to medical research. A paper version of the selfreport questionnaire was posted to 200 registrants over the age of 18 on the HMRI research register. Psychology students over the age of 18 were recruited via an advertisement similar to the Facebook advertisement on the University of Newcastle School of Psychology’s online experimental management system, which allows students to participate in research projects to earn partial course credit. Psychology students completed the online self-report questionnaire hosted by Zoomerang.com. All participants were able to enter into a draw for the chance to win the iPod prize. Analysis Missing values analyses were conducted on the survey data. We used the expectation maximisation (EM) method to replace missing values. EM has been criticised for reducing SE rates, and P values should be interpreted with caution, but EM is still recommended as the simplest and most reasonable approach for imputing missing values [34]. To assess the comparative reliability of testing hazardous and harmful substance use among people recruited via Facebook and other methods, we examined the psychometrics of the three substance use measures (FTND, AUDIT and CUDIT) for each group. Cut-off scores ≥ 8 were chosen as indicators of © 2015 Australasian Professional Society on Alcohol and other Drugs
4
L. K. Thornton et al.
AUDIT, CUDIT total scores and psychiatric symptom measures (DASS-21).
high-severity tobacco, alcohol and cannabis use on the FTND, AUDIT and CUDIT. Scores of 8 or greater on the AUDIT and CUDIT have been shown to be a reliable indicator of hazardous and harmful use and possible alcohol/cannabis dependence while scores of 8 or greater on the FTND indicate very high dependence. We conducted Pearson’s χ2 analyses among participants reporting any substance use, to assess whether the sample groups differed significantly on the proportion of participants who were identified as high-severity. To investigate if, and how, participants recruited via Facebook differed from those recruited via other methods we combined the university student sample with the HMRI sample for comparisons with the Facebook sample. We compared our Facebook and non-Facebook sample groups on the key demographic factors of sex and age. A one-way analysis of variance (anova) was conducted to compare participant ages across recruitment sources. A Pearson’s χ2 analysis was used to compare sex among the samples. Similarly, the percentages of participants reporting current (within the last week) and past use of tobacco, alcohol and cannabis, and endorsing a history of mental health issues were compared between the Facebook and nonFacebook samples using Pearson’s chi-square analyses. Next we conducted general linear modelling to test how the sample groups compared on the FTND,
Results The study’s Facebook advertisement had 4 106 729 impressions over the course of 1 month.This resulted in a total of 2220 clicks (0.054% click-through rate), of whom 553 went on to complete the self-report questionnaire. Twenty-nine (5.24%) participants were excluded because of excessive missing values leaving a total of 524 participants. Over the course of the campaign, the cost-per-click ranged between $0.30 and $0.63 (mean = $0.44). Overall, the campaign cost $975.83 or $1.86 per participant recruited via Facebook. In addition to the participants recruited via Facebook, 115 HMRI research registrants and 319 psychology students returned the self-report questionnaire completed. Seven (6.08%) HMRI registrants and 9 (2.82%) psychology students were excluded because of excessive missing values, leaving a total of 418 nonFacebook participants. Participant characteristics are displayed in Table 1. We first compared the psychology student and HMRI samples on the FTND, AUDIT, CUDIT and DASS21 subscales, finding no statistically significant
Table 1. Participant characteristics Facebook
Non-Facebook
Male N Age Range Mean (SD) ATSI Substance use Ever smoked Smoked in last week Ever drank Drank in last week Ever used cannabis Used cannabis last week History of mental health issues Depression Anxiety Psychotic disorder Eating disorder Other
352
Female 172
Male 89
Female 329
18–73 30.68 (12.44) % 2.8
18–62 32.72 (12.73) % 4.1
18–86 33.16 (18.87) % 3.4
18–80 29.44 (14.98) % 3.3
88.6 61.6 85.8 75.0 65.1 46.3
77.9 56.8 89 58.6 58.1 26.0
68.5*** 21.7*** 95.5* 72.0 42.7*** 23.7**
51.4*** 21.4*** 93.9* 67.1 35.6*** 7.0***
20.5 12.5 3.2 0.0 6.5
36.6 20.3 1.8 3.5 11.6
20.2 10.1 1.1 1.1 3.4
19.6*** 14.4 0.9 4.3 3.0***
***P < 0.001; **P < 0.01; *P < 0.05. All characteristics are self-reported. ‘Other’ mental health issues listed by participants were primarily bipolar disorder, attention deficit/hyperactivity disorder and borderline personality disorder. ATSI, Aboriginal and Torres Strait Islanders. © 2015 Australasian Professional Society on Alcohol and other Drugs
Recruiting via Facebook
5
Table 2. Drug use scale psychometrics among participants reporting tobacco, alcohol or cannabis use Measure/group AUDIT Facebook Non-Facebook FTND Facebook Non-Facebook CUDIT Facebook Non-Facebook
N
Range
Mean
SD
Alpha
S–B
455 394
0–38 0–30
9.13 6.84
7.52 5.41
0.87 0.80
0.85 0.80
446 230
0–10 0–9
2.45 1.26
2.59 1.63
0.82 0.78
0.91 0.86
329 155
0–39 0–29
8.35 4.23
8.34 4.23
0.88 0.84
0.86 0.90
Facebook = Facebook sample, non-Facebook = university student and community volunteers. Alpha, Cronbach’s alpha; AUDIT, Alcohol Use Disorders Identification Test; CUDIT, Cannabis Use Disorders Identification Test; FTND, Fagerstrom Test for Nicotine Dependence; S–B, Spearman–Brown prophecy coefficient.
Table 3. Sample comparisons on proportions of substance using participants reporting high-risk substance use Women Measure AUDIT (n = 849) FTND (n = 676) CUDIT (n = 484)
Men χ2
Facebook (%) Non-Facebook (%) 32.7 6.0 24.0
32.7 1.8 4.3
φ
Facebook (%) Non-Facebook (%)
0.00 0.00 3.76* 0.11 18.12*** 0.29
55.3 4.2 42.8
47.1 1.6 15.8
χ2
φ
1.81 .07 0.90 .05 10.00** .19
*P = 0.05; **P < 0.01; ***P < 0.001. Facebook = participants recruited via Facebook, non-Facebook = psychology student and community participants. AUDIT, Alcohol Use Disorders Identification Test; CUDIT, Cannabis Use Disorders Identification Test; FTND, Fagerstrom Test for Nicotine Dependence.
differences (P > 0.05). We therefore concluded that the two groups were comparable and combined them as a non-Facebook sample. Table 2 displays the ranges, measures of centrality, and internal reliability values for each of the three substance use measures. We report both Cronbach’s α and Spearman–Brown prophecy coefficient (a measure of split-half reliability) to provide more comprehensive information on scale performance. The Facebook sample was somewhat larger, and therefore it is not surprising that a broader range of total scores was found in that group. Although all measures demonstrated adequate internal reliability (i.e. Cronbach’s α ≥ 0.70), for both samples, the Facebook sample showed small improvements in internal reliability in five of six comparisons. The mostly improved internal reliability levels, for the Facebook sample, are likely to have been influenced by the broader range of total scores on the three measures. As can be seen in Table 1, the Facebook and nonFacebook samples differed on sex, χ2 (N = 942, degrees of freedom = 1) = 196.61, P < 0.001, Phi = 0.46, but did not differ significantly on age. For this reason, subsequent analyses were conducted separately for
males and females. Significantly more male and female Facebook participants reported current and past use of tobacco and cannabis, while significantly more nonFacebook participants reported past use of alcohol; however, the samples did not differ on current use of alcohol. Women in the Facebook sample were more likely to report a history of depression, and ‘other’ mental health issues than women in the non-Facebook sample. Table 3 shows that the Facebook sample generally captured greater percentages of high-severity users of each substance, when compared with the nonFacebook sample, for each sex. However, only cannabis use showed statistically significant differences, for each sex. Lastly, we conducted general linear modelling to determine if the Facebook sample differed from the non-Facebook sample on FTND, AUDIT and CUDIT total scores and DASS21 subscales. As the Facebook and non-Facebook samples were previously found to differ on sex, to better understand how these samples compared on our key variables of interest, we conducted the following analyses separately by sex. Multivariate anova showed significant group effects on the © 2015 Australasian Professional Society on Alcohol and other Drugs
6
L. K. Thornton et al.
Table 4. Results of MANOVAs comparing Facebook and non-Facebook samples on drug use and psychiatric symptoms Women Facebook (n = 89) Variable Drug use AUDIT FTND CUDIT DASS21 Depression Anxiety Stress
Men
Non-Facebook (n = 99)
Mean
SD
Mean
SD
7.82 2.37 6.48
6.99 2.50 7.28
8.67 1.57 3.82
5.40 1.97 3.14
5.61 3.55 5.50
4.75 3.83 4.26
3.54 2.80 4.54
4.02 3.51 3.80
Facebook (n = 219) F
Non-Facebook (n = 35)
d
Mean
SD
Mean
SD
F
d
0.87 6.07* 11.00**
0.14 0.36 0.48
10.90 2.54 9.18
7.65 2.62 8.40
10.80 1.17 5.43
6.78 1.54 6.21
0.01 9.03** 6.40*
0.01 0.55 0.46
26.38*** 4.88**** 6.60*
0.48 0.21 0.24
5.10 3.43 5.34
4.34 3.28 3.80
3.71 2.88 4.38
4.18 3.59 3.84
7.44* 1.96 4.46****
0.32 0.16 .25
*P < 0.05; **P < 0.01; ***P < 0.001 (Bonferroni corrected, α/3); ****P < 0.05, (uncorrected). Facebook = participants recruited via Facebook, non-Facebook = psychology student and community participants Only participants who reported use of tobacco, alcohol and/or cannabis in the past 6 months completed the FTND, AUDIT and/or CUDIT. AUDIT, Alcohol Use Disorders Identification Test; CUDIT, Cannabis Use Disorders Identification Test; DASS, Depression Anxiety Stress Scales; FTND, Fagerstrom Test for Nicotine Dependence; manova, multivariate analysis of variance.
substance use and DASS-21 subscales (P < 0.001), for each sex. As the sample groups showed statistically significant differences on these combined variables, we conducted follow-up anovas to assess each measure separately. Table 4 illustrates that for both men and women, the Facebook sample did report significantly more harmful use of tobacco and cannabis. However, the samples did not differ on reported harmfulness of alcohol use. The Facebook sample reported significantly more severe depressive symptoms. Women, recruited through Facebook, also reported significantly higher stress. Discussion This study is one of the first to examine and describe the process of recruiting participants to addiction research via Facebook. At a cost of $1.86 per participant it demonstrates that recruiting via social media sites is a potentially efficient and cost-effective method for recruiting large numbers of participants, including traditionally hard-to-access groups of people, to addiction research. This study set out to identify the type of person who might participate in addiction research via Facebook. We found that people of a wide range of ages, but mostly men, responded to our Facebook advertisement. With a mean age of 31, results show that this approach may not only be useful for recruiting young people to research, but may also be helpful when aiming to recruit older adults as well. We also found that a large proportion of Facebook participants reported tobacco © 2015 Australasian Professional Society on Alcohol and other Drugs
and cannabis use. However, future studies should examine the relationship between self-reported substance use (as reported via Facebook, for example), with objective measures of substance use (e.g. bloods, urine, collateral reports) to establish the reliability of reporting via this source. This research suggests, however, that social media sites, such as Facebook, may be an effective way to gain access to large numbers of substance users who are willing to participate in research. This study also suggests Facebook may be a useful source for recruiting participants with mental health issues. Previous research examining online recruitment and data collection suggests this approach may reduce bias in response to sensitive topics such as substance use and mental health [19]. We therefore hypothesised that compared with participants recruited via more traditional methods, more participants recruited via Facebook would report use of tobacco, alcohol or cannabis and a history of a mental health issues. Similarly, we also hypothesised that these participants would report more harmful use of these substances and more severe psychiatric symptoms. These hypotheses were, for the most part, supported by the current results. Greater percentages of Facebook participants reported past and current use of tobacco and cannabis and a history of depression. Current use of alcohol did not differ between the groups. Similarly, significantly more Facebook participants reported high-severity cannabis use. Mean FTND and CUDIT scores were also significantly higher among Facebook participants, indicating more harmful use of tobacco and cannabis among
Recruiting via Facebook
Facebook participants. While all DASS21 subscale scores were in the expected direction, indicating greater psychiatric distress among participants recruited via Facebook, only depressive symptoms and stress among women were significantly more severe in the Facebook sample compared with the non-Facebook sample. These results suggest that while samples recruited via social media sites such as Facebook might differ from those recruited via more traditional methods, these differences highlight the potential usefulness of this approach for addiction research. The results also question the utility of combining samples recruited in these different ways, as they appear not to be from the same population. Recruiting online and via social media sites may allow researchers to easily access larger numbers of high-severity substance users, and those with comorbid mental health issues, than is usually possible via traditional methods. A number of criticisms often directed at studies employing Internet recruitment methods could also be levelled at the current study. Concerns are often raised regarding the quality of data collected in Internet studies. In the current study similar percentages of participants were excluded from all three recruitment sources because they did not adequately complete the self-report questionnaire (Facebook: 5.24%; HMRI: 6.08%; psychology students: 2.82%).This suggests that it may be possible to collect equivalent quality data via Facebook in comparison with other recruitment sources. Yet, this study did not directly test the genuineness of participant response; that is, the survey did not make use of questions to elucidate response truthfulness, and did not control for participant response patterns (e.g. always responding A in a multiple choice question). Future research in this area would benefit by directly testing the genuineness of participants responses across recruitment methods. Another limitation of the current study was that because comparing Facebook with other recruitment methods was not an aim of the original study, we did not track the costs associated with the other recruitment methods used. Future research should aim to directly compare the cost-effectiveness of various recruitment methods. Additionally, it should be noted that the samples used as a comparison with the Facebook sample, psychology students and volunteers from community research databases are unlikely to be fully representative of the general population. However, these two groups are common sources of participants in behavioural research [35]. A common criticism of the use of Internet advertising in research is the frequently low click-through rate [2]. In the current study, the click-through rate was 0.054%. Similarly, low click-through rates have been
7
found in other studies using comparable methods [2,3]. A number of researchers have been critical of these low click-through rates, comparing click-through rates with response rates of traditional recruitment methods such as mail-outs. This is an inappropriate comparison with infer, as mail-outs directly approach people to participate in research. A more appropriate comparison with make would be with the recruitment of participants through indirect or passive advertising via traditional media such as TV, radio and print media. The Internet is a quickly changing arena, and recent changes within this sphere may affect the continuing relevance of the current findings. In this study, and previous studies, it is proposed that the greater reporting on sensitive topics observed may have been due to the perceived privacy associated with online research. However, starting in June 2013, Edward Snowden revealed information that showed that many governments, including Australia, are conducting large-scale digital surveillance on their own citizens, and those of other countries [36,37]. It is unclear if events have eroded the perception of anonymity and privacy online to a point where social desirability biases would now influence participants responses to sensitive topics more [36]. Other changes include those involving the way in which advertising campaigns can be developed and disseminated on Facebook. Facebook has made new tools available that allow advertisers to target particular audiences more specifically and provide detailed information about the potential reach of advertisements [27]. For example advertisements can be targeted according to a range of users’ behaviours such as: online spending activity, the device or programme they use to access Facebook, and even if their Facebook activity suggests they are planning to travel or have recently returned from a trip [27]. Despite these limitations, this paper highlights the potential of social media sites, such as Facebook as a recruitment source for addiction research. It is important to note, however, that not all Facebook advertising campaigns are as successful as the campaign employed in the current study, and not all research topics and questions will be appropriate for this method of recruitment. The current study, for example, gave participants the chance to win an iPod, which may have contributed significantly to the success of the campaign. Indeed, in a meta-analysis of online surveys Goritz [38] found that offering an incentive significantly increased the odds of a person responding to, and completing, an online survey. The current study also likely oversampled substance users as a result of the advertising procedure used by Facebook, which meant the advertisement would have been more likely to appear on the pages of people whose activity contained mentions of drugs and alcohol [39]. © 2015 Australasian Professional Society on Alcohol and other Drugs
8
L. K. Thornton et al.
Conclusion Facebook, and other social media sites, have the potential to play an important role in accessing substance users and engaging them with research and potentially treatment, as well as increasing our understanding of substance use and addictive behaviours. This research found that large numbers of cannabis users were willing to participate in research via Facebook. As computerdelivered cannabis use treatment has been found to be effective [40], Facebook and other social media sites could be an effective way to access and engage cannabis users with effective treatment. These sites could also be an effective way to directly reach and engage young people with addiction treatment and research. This approach could address the potential barriers of obtaining parental consent and fear that parents will find out about any substance use the young person reports. Finally, social media sites, like Facebook could also facilitate rapid research regarding new and emerging substances. Acknowledgements This research was supported by Dr Thornton’s Australian Postgraduate Award. References [1] Balfe M, Doyle F, Conroy R. Using Facebook to recruit young adults for qualitative research projects: how difficult is it? Comput Inform Nurs 2012;30:511–15. [2] Fenner Y, Garland SM, Moore EE, et al. Web-based recruiting for health research using a social networking site: an exploratory study. J Med Internet Res 2012;14: e20. [3] Ramo DE, Prochaska JJ. Broad reach and targeted recruitment using Facebook for an online survey of young adult substance use. J Med Internet Res 2012;14:e28. [4] Temple EC, Brown RF. A comparison of Internet-based participant recruitment methods: engaging the hidden population of cannabis users in research. J Res Pract 2011;7(2), Article D2. Retrieved [16th June 2015], from http://jrp.icaap.org/index.php/jrp/article/view/288/247 [5] Deady M, Kay-Lambkin F, Thornton L, Baker A, Teesson M. Social influence, addictions and the internet: the potential of Web 2.0 technologies in enhancing treatment for alcohol/other drug use problems. Addiction Research and Therapy 2012;8:2. [6] Freeman B, Chapman S. Gone viral? Heard the buzz? A guide for public health practitioners and researchers on how Web 2.0 can subvert advertising restrictions and spread health information. J Epidemiol Community Health 2008; 62:778–82. [7] Boyd DM, Ellison NB. Social network sites: definition, history and scholarship. J Comput-Mediat Comm 2008; 13:210–30. [8] Alexa.com. 2014 Alexa.com. Available at: http://www .alexa.com/siteinfo/facebook.com (accessed April 2014). © 2015 Australasian Professional Society on Alcohol and other Drugs
[9] Hamptom K, Sessions G, Rainie L, Purcell K. Social networking sites and our lives. Washington D.C.: PEW Internet & American Life Project, 2011. [10] Farmer AD, Bruckner Holt CEM, Cook MJ, Hearing SD. Social networking sites: a novel portal for communication. Postgrad Med J 2009;85:455–9. [11] Gilligan C, Kypri K, Bourke J. Social networking versus Facebook advertising to recruit survey respondents: a quasi-experimental study. JMIR Res Protoc 2014;3: e48. [12] Hernandez-Romieu AC, Sullivan PS, Sanchez TH, et al. The comparability of men who have sex with men recruited from venue–time–space sampling and Facebook: a cohort study. JMIR Res Protoc 2014;3:e37. [13] Ramo DE, Rodriguez TM, Chavez K, Sommer MJ, Prochaska JJ. Facebook recruitment of young adult smokers for a cessation trial: methods, metrics and lessons learned. Internet Interv 2014;1:58–64. [14] Akard TF, Wray S, Gilmer MJ. Facebook advertisements recruit parents of children with cancer for an online survey of web-based research preferences. Cancer Nurs 2015;38: 155–61. [15] Schumacher KR, Stringer KA, Donohue JE, et al. Social media methods for studying rare diseases. Pediatrics 2014; 133:e1345–53. [16] Lohse B, Wamboldt P. Purposive Facebook recruitment endows cost-effective nutrition education program evaluations. JMIR Res Protoc 2013;2:e27. [17] O’Conner A, Jackson L, Goldsmith L, Skirton H. Can I get a retweet please? Health research recruitment and the Twittersphere. J Adv Nurs 2014;70:599–609. [18] Newman JC, Des Jarlais DC, Turner CF, Gribble J, Cooley P, Paone D. The differential effects of face-to-face and computer interview modes. Am J Public Health 2002;92: 294–7. [19] Crutzen R, Göritz AS. Social desirability and self-reported health risk behaviors in web-based research: three longitudinal studies. BMC Public Health 2010;10:720. [20] Joinson AN. Knowing me, knowing you: reciprocal selfdisclosure in Internet-based surveys. Cyberpsychol Behav 2001;4:587–91. [21] Kogan N. Risk taking: a study in cognition and personality. New York: Holt, Rinehart &Winston, 1964. [22] Joinson AN, Woodley A, Reips UD. Personalization, authentication and self-disclosure in self-administered Internet surveys. Comput Human Behav 2007;23:275– 85. [23] Ellis LA, Collin P, Davenport TA, Hurley PJ, Burns JA, Hickie IB. Young men, mental health, and technology: implications for service design and delivery in the digital age. J Med Internet Res 2012;14:e160. [24] Hilton S, Smith E. ‘I thought cancer was one of those random things. I didn’t know cancer could be caught . . .’: adolescent girls’ understandings and experiences of the HPV programme in the UK. Vaccine 2011;29:4409–15. [25] Chu JL, Snider CE. Use of a social networking web site for recruiting Canadian youth for medical research. J Adolesc Health 2013;52:792–4. [26] Close S, Smaldone A, Fennoy I, Reame N, Grey M. Using information technology and social networking for recruitment of research participants: experience from an exploratory study of pediatric Klinefelter syndrome. J Med Internet Res 2013;15:e48. [27] Facebook 2013 Facebook.com. Available at: http://www .facebook.com (accessed January 2013).
Recruiting via Facebook
[28] Brickman Bhutta C. Not by the book: Facebook as a sampling frame. Sociol Methods Res 2012;41:57–88. [29] Thornton LK, Baker AL, Johnson MP, Lewin TJ. Perceived risk associated with tobacco, alcohol and cannabis use among people with and without psychotic disorders. Addict Behav 2013;38:2246–51. [30] Heatherton TF, Kozlowski LT, Frecker RC, Fagerstrom KO. The Fagerstrom Test for Nicotine Dependence: a revision of the Fagerstrom tolerance questionnaire. Br J Addict 1991;86:1119–27. [31] Saunders JB, Aasland OG, Babor TF, de la Fuente JR, Grant M. Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption—II. Addiction 1993;88:791–804. [32] Adamson SJ, Sellman JD. A prototype screening instrument for cannabis use disorder: the Cannabis Use Disorders Identification Test (CUDIT) in an alcohol-dependent clinical sample. Drug Alcohol Rev 2003;22:309–15. [33] Lovibond SH, Lovibond PF. Manual for the depression anxiety stress scales, 2nd edn. Sydney: Psychology Foundation, 1995.
9
[34] Tabachnick BG, Fidell LS. Using multivariate statistics, 5th edn. Boston: Pearson Education, 2007. [35] Samuels D, Zucco Cr. Using Facebook as a Subject Recruitment Tool for Survey-Experimental Research (August 2013). 2013. Available at SSRN: http://ssrn .com/abstract=2101458 or http://dx.doi.org/10.2139/ssrn .2101458. [36] Lyon D. Surveillance, Snowden, and big data: capacities, consequences, critique. Big Data & Soc 2014;July– September:1–13. [37] Dorling P. Top-secret documents reveal Australia targeting Indonesia, South Pacific mobile phone networks. The Sydney Morning Herald. 2015;2015. [38] Goritz AS. Incentives in web studies: methodological issues and a review. Int J of Internet Science 2006;1:58–70. [39] Facebook. 2014. About Facebook ads. Available at: https:// www.facebook.com/about/ads/ (accessed 16 June 2015). [40] Kay-Lambkin F, Baker A, Lewin T, Carr V. Computerbased psychological treatment for comorbid depression and problematic alcohol and/or cannabis use: a randomized controlled trial of clinical efficacy. Addiction 2009;104: 378–88.
© 2015 Australasian Professional Society on Alcohol and other Drugs
